Alright I'm back with some interesting results that confirm my suspicions (actually meant 300/30 Mbit btw)
* All VM tests (xp/w7/w10/ubuntu) showed similar results around 0.9..1MB/s upload, except when using a UK VPN: ICS was again slow. * Win XP on a netbook on wifi (Asus eee901): ICS: 65.8s. SYN: 6.5s ** Fresh Win7 on my PC: * - ICS: Upload Duration: 34.39 secs/132,417 chars/sec - SYN: Upload Duration: 2.03 secs/Speed: 2,241,135 chars/sec ** Fresh Win10 on my PC** - ICS: Upload Duration: 1.88 secs/ 2,427,498 chars/sec - SYN: Upload Duration: 2.02 secs/Speed: 2,257,802 chars/sec ** Fresh Ubuntu16 on my PC: (wine)* - ICS: Upload Duration: 1.41 secs/Speed: 3,236,664 chars/sec - SYN: Upload Duration: 1.41 secs/Speed: 3,234,365 chars/sec - Firefox: Upload Duration: 1.35 secs/Speed: 3,385,863 chars/sec As for the GET, the 2 ICS links seem a bit slower than the IIS. Did a GET test on a 5MB file: http://download.thinkbroadband.com/5MB.zip Both ICS and SYN took around 4.1-4.3s, no issues with http downloads I think. My conclusion so far: Up to Windows 7, ICS is interacting with Winsock in a manner that causes such extreme low post speeds, and it seems to be "fixed" in Windows 10 (possibly since W8?) The VM tests showed similar results on all platforms, most likely due to ~negating the "wrong" way ICS is interacting with the server, possibly something like.. VM->ICS posts data to locahost->VMware bridge with host captures this data and sends it a different way to the actual server, negating the root ICS issue, sending data its own way no matter the guest platform. Also interesting how much faster the Linux tests were :O Did you compile you application with "release" config, and optimization and > without any debugging tool enabled such as madExcept or FastMM debug mode ? > Yeah, you can check in the test app I linked a few posts earlier, release/no exception handler/no FastMM. edit: I'm about to install a fresh Win8 now, will post back. On Wed, Apr 27, 2016 at 2:05 PM, Angus Robertson - Magenta Systems Ltd < an...@magsys.co.uk> wrote: > > So I'm kinda baffled by these results, not sure what to make of it. > > I do know from long experience that ruling out a single PC is the first > thing I do when testing stuff like this. But even then, sometimes > results simply can never be explained. > > I have speed test downloads from my public servers every morning, and > have known for a long time that HTTP is sometimes slower than it should > be, although today it seems rather good. I have two broadband > connections, 40/10 and 152/12 but use the former for most speed tests, > although I did download 10.1 Berlin ISO at 100Mb last week which would > excellent. > > You have only been doing uploads so far, but I will be mainly testing > downloads when I fix the buffers, using: > > http://www.telecom-tariffs.co.uk/testing/speed50meg.zip > http://www.magsys.co.uk/download/testing/speed50meg.zip > http://www1.telecom-tariffs.co.uk/testing/speed50meg.zip > http://www1.magsys.co.uk/download/testing/speed50meg.zip > > The first URL is ICS, second IIS, www is Windows 2012 R2, www1 is 2008. > As I said earlier, these servers are behind a slowish firewall since I > get continual hacking attempts and Windows servers are so open to > attack, the set-up was never designed for high speed testing. > > Angus > > -- > To unsubscribe or change your settings for TWSocket mailing list > please goto http://lists.elists.org/cgi-bin/mailman/listinfo/twsocket > Visit our website at http://www.overbyte.be > -- To unsubscribe or change your settings for TWSocket mailing list please goto http://lists.elists.org/cgi-bin/mailman/listinfo/twsocket Visit our website at http://www.overbyte.be
